alibabacloud-idaas-create-saml-idp

Pass

Audited by Gen Agent Trust Hub on Nov 2, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted SAML metadata in XML format sourced from external URLs or local files. This creates a surface for indirect prompt injection where malicious instructions could be embedded in the metadata.
  • Ingestion points: SAML metadata is fetched via the fetch_metadata function or read from local files in scripts/idaas_saml_idp.py.
  • Boundary markers: The script parses the XML into structured JSON, and the agent is explicitly instructed to present a configuration plan for user confirmation before performing any cloud write operations.
  • Capability inventory: The skill can execute Alibaba Cloud CLI commands for identity provider management and write state/constraint files to the local workspace.
  • Sanitization: The script performs URL validation (SSRF protection), enforces HTTPS, and strictly verifies that only public certificates are imported, rejecting any input containing private key material.
  • [DYNAMIC_EXECUTION]: The main script utilizes the xml.etree.ElementTree library to parse SAML descriptors. This standard library is susceptible to XML External Entity (XXE) attacks when processing untrusted XML. Additionally, a regression test script (scripts/test_run32294_regressions.py) uses __import__ to dynamically load modules, though this is confined to the test environment.
  • [COMMAND_EXECUTION]: The skill executes the aliyun CLI and curl using subprocess.run. These calls are safely implemented using argument lists instead of shell strings, which effectively prevents shell injection vulnerabilities.
  • [EXTERNAL_DOWNLOADS]: The skill downloads SAML metadata from user-provided URLs. It includes SSRF mitigation by validating that URLs do not point to localhost or private network addresses, although these restrictions can be bypassed if the user explicitly enables insecure testing mode.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 2, 2026, 06:38 AM